OpenAI Claims AI Solved 90-Year Math Problem, But Credit Dispute Clouds Announcement

·3 views

On September 8, OpenAI announced that an internal AI system — a swarm of roughly 10,000 concurrent agents running on an unreleased model — solved the Navier-Stokes existence and smoothness problem, one of seven unsolved Millennium Prize Problems carrying a $1 million award. The system ran for about 88 hours, exchanged 2.7 million messages, and cost an estimated $15 million in compute before producing a Lean-verified proof of a fluid 'blow-up' singularity. Hours before OpenAI's announcement, mathematicians Tristan Buckmaster of NYU and Levent Alpoge of Anthropic independently published their own Lean-verified blow-up result for the related Euler equations. Buckmaster subsequently alleged that OpenAI may have had prior knowledge of his work through its Codex tools, which he had used during research, and accused the company of shifting its account of how the AI result was reached. He further claimed OpenAI proposed announcement arrangements that would have omitted Alpoge from credit, reportedly due to his affiliation with rival Anthropic.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Why Amazon Dropped MOBI for EPUB and How Browsers Can Convert Files Privately

Amazon formally phased out MOBI support for its Send to Kindle service in late 2022, pushing users toward the EPUB format instead. MOBI, originally created by Mobipocket in 2000 and later acquired by Amazon in 2005, lacked modern typography features such as CSS layouts, SVG support, and dynamic dark mode styling. EPUB, standardized by the W3C and IDPF, is built on HTML5, CSS3, and XML, making it far better suited to contemporary Kindle rendering engines. When users sought alternatives to convert their existing MOBI libraries, many turned to online tools that upload files to third-party servers, raising serious privacy concerns around data retention and exposure. Modern browser technologies like WebAssembly and FileReader APIs now enable fully client-side MOBI-to-EPUB conversion, meaning files never leave the user's device.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Deploy a Web Project to Tencent EdgeOne Makers via GitHub

Tencent EdgeOne Makers offers a streamlined way to deploy web projects by connecting directly to a GitHub repository, eliminating the need for manual file uploads or server configuration. Developers can sign in to the EdgeOne Makers dashboard, authorize access to their GitHub account, and select the desired repository for deployment. The platform automatically detects common frameworks like React, Vue, or Vite and suggests appropriate build settings. Once deployed, EdgeOne Makers provides a live URL to access the website. Any subsequent code changes pushed to the connected GitHub branch can automatically trigger a new deployment, keeping the live site up to date.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Agent benchmarks measure memory-less models, missing a key performance variable

A critique published on DEV Community argues that AI coding-agent benchmarks are structurally flawed because they always start models with empty memory, ignoring how accumulated project context affects real-world performance. The argument uses Engrim, an open-source SQLite-based persistent memory engine for tools like Claude Code and Cursor, as evidence that cross-session state is a meaningful and underexplored performance axis. Current evals deliberately reset memory between runs to ensure reproducibility, but this means the benchmarked agent bears little resemblance to one that has ingested months of project decisions. The author proposes a fix that preserves reproducibility: run each task under three conditions — empty context, small curated memory, and large memory — and measure how scores shift across conditions. Until benchmarks account for persistent state, their results represent only a performance floor, not a reliable predictor of production behavior.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Fluent but Wrong AI Code Reviews Are More Dangerous Than Obvious Errors

A analysis published on DEV Community argues that the most dangerous failure in AI code review tools is not a missed bug but a structurally incorrect review that appears credible and trustworthy. Drawing on a database-recovery case study by Oskar Gross at Glazer, the piece illustrates how readable output with corrupted meaning — like CSV values under wrong headers — is worse than an obvious error because it goes unquestioned. The author warns that many AI review tools assess only whether code reads well, rather than validating it against the broader codebase, its types, contracts, and callers. A tool that produces fluent but structurally blind feedback can cause developers to stop scrutinizing changes precisely where scrutiny is most needed. The piece urges teams to benchmark AI review tools on worst-case scenarios — specifically the subset of changes where the tool gives plausible but wrong feedback — rather than relying on average bug-catch rates.

0 comments Read more at DEV Community